cancel
Showing results for 
Search instead for 
Did you mean: 

Modeling of many products - How to efficiently create models

Modeling of many products - How to efficiently create models

Hi teams,

I'm developing prediction models of sales order.

I have around 200 products (SKUs) and currently I'm creating 1 model per 1 product.

The prediction result looks good for backtest, but considering deploy it to production, I think 200 models would be too much to be managed.

 

In this case, I would have to develop a model including multiple products.

What do you think would be the best criteria for grouping products?

 

For example:

1) Sales volume

Model A: 100-1000 pcs per month

Model B: 1000-5000 pcs per month

Model C: 5000-10000 pcs per month

 

2) Sales trend 

Model A: Stable linear trend

Model B: Trend with seasonality

Model C: Volatile

 

I assume combination of 1) and 2) is better but it would be a bit difficult to correctly determine 2).

 

Or rather to stick with 1 product 1 model approach?

(200 models are a regular case in DataRobot?)

 

I really appreciate the feedback from your expertise and experience!

Labels (3)
0 Kudos
0 Replies